FIG. 1 (Prior Art) is a diagram of an exemplary Internet structure 100. Internet structure 100 includes numerous servers 101-110 (each designated with an "S") interconnected by high-speed connections 111-119. The high-speed connections are of several different types: for example, connection 113 is a T1 connection, connection 115 is a high-speed fiber-optic connection, part of the so-called "backbone" of the Internet, and connection 118 is a lower-speed ISDN connection. Internet structure 100 also includes numerous clients (each designated with a "C") connected to the servers. Most of the clients are connected to the Internet via an Internet Service Provider, or "ISP." For example, client 120 is connected to the Internet via ISP1 server 101.
ISPs maintain servers on the Internet and sell Internet access to individual subscribing clients. A subscribing client typically gains access to the Internet using a modem to call the ISP. The client typically dials a local telephone phone number to establish a connection to the Internet via the ISP by connecting to the ISP's point of presence (POP). A local area network (LAN) of clients can also be connected to the Internet, either via a dial-up connection or a permanent connection. For example, LAN 123 is connected to the Internet via one of the nodes 104 on LAN 123 that happens to be a server.
In the example, server 110 is the server of a large bookstore that advertises and sells books over the Internet, for example by maintaining a web site. The customers of the bookstore access the Internet as clients 120-122. Customer clients 120 and 122 may, for example, be individuals who buy books and access the Internet from their respective homes in California and New York. The bookstore server 110 may, for example, be located in Illinois.
Companies that advertise on the Internet recognize the importance of providing customers and potential customers with a pleasurable shopping or browsing experience. On the Internet, a pleasurable experience generally requires that users can easily establish and maintain fast, reliable connections, and that the various elements of the site are available. Users may have negative experiences when connecting to a web site using particular ISPs or local dial-up connections. These negative experiences can result in low sales figures in some areas. For example, potential customers in San Francisco may have difficulty accessing bookstore 110 due to local Internet infrastructure limitations. Companies that advertise on the Internet therefore spend large sums of money trying to determine whether customers and potential customers in various parts of the world can easily access, explore, and interact with company web sites.
FIG. 2 (Prior Art) is a diagram of an Internet structure 200 in which a client 124 provides a measure of user experience for specified Internet sites. Client 124 monitors the quality of Internet connections to Internet sites and sells the gathered information to companies that advertise on the Internet.
Client 124 dials out from a single location and accesses a target Internet site (for example, bookstore server 110) via a number of ISPs 101, 103 and 109. Client 124 then records performance data indicative of the quality of connections to the target site. Unfortunately, the different lengths of the long-distance connections 124A, 124B and 1245C from the single location of client 124 to the various ISPs 101, 103 and 109 affects connection quality. In contrast, the customers 120, 121 and 122 of the bookstore are connected to ISPs via local dial-up telephone connections 120A, 121A and 122A. The performance data collected by client 124 at the single location is therefore not necessarily representative of a typical user's experience.
FIG. 3 (Prior Art) is a simplified diagram of an Internet structure 300 in which a "distributed" system of data-gathering computers 125-127 provides a measure of user experience for specified Internet sites. The data-gathering computers 125-127 are linked to the Internet at various geographically separated locations via dedicated connections 128-130 (not dial-up connections). The performance data collected is not representative of a typical user's experience for at least two reasons. First, the data-gathering computers are connected to the Internet via dedicated connections rather than dial-up connections. Second, the software used to access the target site is different than the commercially available browser software used by typical users. Moreover, the systems of FIGS. 2 and 3 do not measure some performance characteristics that are useful in assessing user experience; for example, neither system measures the time required for a user to download a given web page.